Chair objectives
The main objectives of this Chair is to develop and to put in action upstream applied mathematical
tools (optimization tools, non Euclidean geometry, global sensitivity analysis, random matrices, ..),
for the understanding of machine learning algorithms.
A special emphasis is put on the strong
interactions with our industrial partners. Our wish is to work in a virtuous loop, feeding our
upstream researches in applied mathematics by solving concrete industrial problems proposed by
our industrial collaborators. The addresed problems range from explicability (using for example tools
of global sensitivity analysis), the study of neural nets (using for example random matrices or
continuous embedding) to the design of new algorithms (using geometric tools).
Programs: Acceptable and certifiable AI
Themes: learning with little of complex data, fair learning, AI and physical models
Chair holder:
Fabrice Gamboa, Institut de Mathématiques de Toulouse / https://www.math.univ-toulouse.fr/~gamboa
- Senior collaborating researchers :
Agnès Lagnoux (UT2J, IMT), Clément Pellegrini (UT3, IMT) - PhD students
Jérôme Stenger (CIFRE EDF. 2017-2020, Advisor F. Gamboa and M. Keller): Optimal uncertainty quantification of a risk measurement from a computer codeEva Lawrence (CEA grant, 2016-2020, Advisors F. Gamboa and T. Klein): Functional reconstruction and uncertainty analysis in an inverse chemical thermodynamics problem
Louis Berry (CEA grant, 2019-2022, Advisors F. Gamboa and A. Marrel): Computing data assimilation of a nuclear engine simulator,
Faouzi Hakimi (CEA grant, 2019-2022, Advisors F. Gamboa and A. Marrel): Sensitivity, and data assimilation of a nuclear accident simulator
Clément Benesse (ENS Lyon ANITI, 2019-2022, Advisors F. Gamboa and J.M. Loubes): Bridging fairness and sensitivity analysis
Virgile Foy (ANITI SAFRAN, 2020-2023, Advisors F. Gamboa and R. Chhaibi ): Machine learning for the automatic generation of engine blades
Marouane Il Drissi (CIFRE EDF, 2020-2023, Advisors F. Gamboa and J.M. Loubes): Game theory, sensitivity analysis and perturbations for machine learning interpretability,
Ismael Khalfaoui (ANITI, 2020-2023, Advisors F. Filbet and T. Pellegrini): Deep neural networks using expanded convolution layers
Co-organisation du workshop SAIS 2020, autour des Statistiques et l’ Intelligence Artificielle pour la Data Science.
Some selected publications
- Daouda, T., Chhaibi, R., Tossou P., & Villani A.-C. (2020) Geodesics in fibered latent spaces: A geometric approach to learning correspondences between conditions. ArXiv:2005.07852 Submitted
- Da Veiga, S., Gamboa, F., Iooss, B. & Prieur C. (2021 to appear) Basics et trends in sensitivity analysis – Theory et practice in R, SIAM.
- De Castro, Y., Gamboa, F., Henrion, D., & Lasserre, J. B. (2021). Dual optimal design et du Christoffel–Darboux polynomial. Optimization Letters, 15(1), 3-8.
- Gamboa, F., Gueneau, C., Klein, T., & Lawrence, E. (2021). Maximum entropy on du mean approach to solve generalized inverse problems with an application in computational thermodynamics. RAIRO-Operations Research, 55(2), 355-393.
- Pellegrini, T., & Masquelier, T. Fast threshold optimization for multi-label audio tagging using surrogate gradient learning. In Proc. ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech et Signal Processing (ICASSP) (pp. 651-655), June 2021
- Pellegrini, T., Zimmer, R., & Masquelier, T. Low-activity supervised convolutional spiking neural networks applied to speech commands recognition. In Proc. 2021 IEEE Spoken Language Technology Workshop (SLT) (pp. 97-103), January 2021
- Stenger, J., Gamboa, F., & Keller, M. (2021). Optimization Of Quasi-convex Function Over Product Measure Sets. SIAM Journal on Optimization, 31(1), 42